The AI Stack Behind the Discovery Engine

At the center of Pinterest’s artificial intelligence strategy is its taste graph, built on hundreds of billions of user interactions over a decade. Every search, save and click adds signal. That signal trains the models that decide what each user sees next.

“When a user knows what they want but cannot quite describe it, an image can do what text cannot,” CEO Bill Ready said on the earnings call.

In Q1, Pinterest extended Pennock, its proprietary generative retrieval system, to serve content globally across all surfaces. Rather than separate models optimized for each surface, Pennock generates personalized results for each user simultaneously, informed by the full depth of their taste and interest history. The launch improved search fulfillment by about 180 basis points and drove a roughly 180 basis point reduction in cost per acquisition and cost per click for advertisers on search surfaces.

Pinterest also updated its search ranking model in Q1, extending user context windows by 30 times. The system now uses up to 16,000 user actions over a two-year period to inform search results. That launch improved search fulfillment by about 70 basis points and saves by about 390 basis points.

The company is also building Canvas; an in-house AI image generation model trained exclusively on Pinterest data. It already supports creative optimization for advertisers, dynamically editing backgrounds and transforming catalog images into lifestyle images. Pinterest says it operates at an order of magnitude lower cost than leading third-party models. The newest version supports real-time image editing in key verticals.

Where the Monetization Gap Lives

Pinterest has delivered five times more clicks to advertisers over the past three years. Revenue hasn’t grown at anything close to that rate. That gap is the business problem Ready is now organized around solving.

“We remain in the early stages of fully monetizing the engagement and commercial intent on our platform,” CFO Julia Donnelly said on the call.

Pinterest Performance Plus, the company’s AI-powered automated ad suite, is the primary vehicle. About 30% of lower funnel revenue now runs through Performance Plus campaigns. Advertisers using it are growing their lower funnel spend at nearly twice the rate of non-adopters. One fine jewelry brand ran a four-week test and saw a 46% increase in return on ad spend and a 62% increase in conversions.

It’s still early. Performance Plus only reached general availability about a year ago. Mid-market, SMB and international advertisers are still a fraction of what they could be.

The larger structural issue is measurement. Pinterest is piloting direct integrations with advertisers’ proprietary measurement systems, letting its AI bidding respond to each advertiser’s specific definition of value, whether that’s lifetime value, profit per order, or something else. In early testing with one advertiser prioritizing lifetime value, the integration drove a 15% to 20% improvement in lifetime value return on ad spend.

Large retailers, Pinterest’s biggest advertiser segment, remained a headwind in Q1. Tariff-related margin pressure is keeping that group cautious. AI-driven bidding optimizations partially offset the drag later in the quarter, but the structural issue hasn’t resolved.

“Of Pinterest’s more than 80 billion monthly searches, half are commercial in nature, whereas ChatGPT’s own data says that only 2% of their prompts are commercial,” Ready said.

What Else Stood Out

• Pinterest acquired TV Scientific in Q1, giving it the ability to extend its taste graph beyond its own platform into connected TV campaigns. Early results with one home furnishings retailer showed a 190% increase in incremental audience reach and a 159% increase in incremental sales. Ready framed CTV as a first move toward monetizing Pinterest’s audience data beyond the Pinterest app.
• Pinterest launched a native A/B testing tool in beta directly in Ads Manager in Q1, letting advertisers run structured tests comparing Performance Plus campaigns to standard ones. The company said early results are strong.
• Gen Z now represents more than 50% of Pinterest’s user base and is its fastest-growing demographic. Ready pointed to Pinterest’s decision to make accounts private by default for users under 16 in 2023 as the catalyst. He said the decision was expected to hurt Gen Z engagement. It did the opposite.
• Pinterest launched a new brand campaign in the U.S. and U.K. earlier this month targeting Gen Z and millennial audiences across television, streaming, cinema, out-of-home and digital channels through the end of the year.

Topline Results and Outlook

Q1 2026 revenue was $1.008 billion, up 18% year over year, or 15% on a constant currency basis, above the high end of guidance. Global MAUs reached 631 million, up 11% year over year, a record for the tenth consecutive quarter. U.S. and Canada revenue was $750 million, up 13%. Europe revenue was $186 million, up 27% reported or 16% constant currency. Rest of world revenue was $72 million, up 59% reported or 50% constant currency. Ad impressions grew 24%; average ad pricing declined 5%.